THE LINUX SERVER TOOLKIT · RELEASE CANDIDATE · 1.0.0 RC1

Know your server.
Measure what
matters.

A clear look at your Linux server. Inspect its resources, run focused performance tests, and keep the results in your hands.

No accountNo automatic uploadsTerminal + JSON

herza-bench — illustrative output

$ curl -fsSL https://herza.sh | bash

HERZA BENCH / 1.0.0-rc.1
By Herza, Ltd.

SYSTEM OVERVIEW
  CPU / RAM / Disk / Virtualization
  Uptime / Load / TCP / Cgroup limits

CPU
  Single worker + parallel workload

STORAGE
  Write x3 / Direct read
  4K IOPS / p95 latency

NETWORK
  Ookla local/best server + 9 international cities
  Upload / Download / Latency

Automatic CLI preparation.

Measured values appear as each test completes.

01 / GET STARTED

Your next benchmark starts here.

Linux · Bash · Python 3.6+ · curl

Run HERZA Bench instantly with one command. One command runs CPU, storage and network tests. Auto mode prepares the official Ookla CLI temporarily when needed and tests the automatically selected server first, followed by 9 international cities. No manual installation. The full network test can take several minutes and use several GiB. Use --local-only for a single local test.

curl -fsSL https://herza.sh | bash

Download and verify this release

Run in an empty directory. The script starts only if the checksum matches.

curl -q -fsS --proto '=https' -O https://herza.sh/releases/1.0.0-rc.1/bench.sh &&
curl -q -fsS --proto '=https' -O https://herza.sh/releases/1.0.0-rc.1/bench.sh.sha256 &&
sha256sum -c bench.sh.sha256 &&
bash bench.sh

Download script ↓Script SHA-256 ↗Download source ↗

Usage & Third-Party Terms: HERZA Bench is provided under the MIT license with usage notes. When you run an Ookla test, HERZA Bench passes acceptance flags for Ookla’s terms. By running that test, you agree to the applicable Ookla EULA and Terms of Use, and acknowledge its Privacy Policy. Run Ookla only if you have the required usage rights. Use --network-engine https or --no-network to run without Ookla.

02 / WHAT WE MEASURE

Useful numbers.
Clear boundaries.

System inventory

CPU model, cache, AES, virtualization, memory and swap, uptime, load, filesystem capacity and TCP congestion. Cgroup limits appear when detected.

Every mode

CPU performance

Integer prime workload: one worker, then up to four workers by default. Primes/s counts prime numbers identified per second. The parallel result combines all workers.

Quick + Full

Disk I/O

1 MiB sequential writes with fsync and direct reads, plus 4 KiB and 1 MiB random I/O at QD1. See median throughput, IOPS and p95 syscall latency. Uses a temporary 64 MiB file.

Default · no fio needed

Network speed

Official Ookla CLI: automatic best-server selection followed by Singapore, Kuala Lumpur, Tokyo, Hong Kong, Paris, Amsterdam, London, Los Angeles, and Dallas. Each test measures upload, download, latency and jitter. Temporary files are created in your working directory, verified with SHA-256, and cleaned up after the run. HTTPS measurements remain an explicit option.

Auto · local/best server first

03 / BUILT FOR TRANSPARENCY

Your server.
Your results.

No HERZA account or automatic package installation. HERZA does not upload report JSON. JSON excludes your public IP and hostname. Test providers see your source IP; their own data policies apply. Use --no-network to skip external measurements.

A measurable step forward.

This revision prepares Ookla CLI automatically for x86_64 and ARM64 Linux. Existing CLI installations can be reused. The default network test passes Ookla license and privacy acceptance automatically. Provider terms and appropriate use rights still apply. Public comparison pages are not included.

Benchmarks consume resources. Use an idle test server for comparisons. Full mode generates disk I/O; network tests consume bandwidth. Compare runs using the same version, settings and endpoint.

Read usage and limitations ↗